Application of Machine Learning for Enhancing Fake Logo Detection

• Olawale J. OLALUYI

  Department of Electrical and Electronic Engineering, Bamidele Olumilua University of Education, Science and Technology, Ikere Ekiti, Nigeria

  Author

• Johnson O. ADEOGO

  Department of Electrical and Electronic Engineering, Bamidele Olumilua University of Education, Science and Technology, Ikere Ekiti, Nigeria

  Author

• Adeniyi O. AJIBOYE

  Department of Electrical and Electronic Engineering, Bamidele Olumilua University of Education, Science and Technology, Ikere Ekiti, Nigeria

  Author

• Mayowa O. ORESELU

  Department of Electrical and Electronic Engineering, Bamidele Olumilua University of Education, Science and Technology, Ikere Ekiti, Nigeria

  Author

• Olarewaju T. OGINNI

  Department of Mechanical Engineering, Bamidele Olumilua University of Education, Science and Technology, Ikere Ekiti, Nigeria

  Author

Machine learning is significant for fake logo detection because it automates accurate identification of counterfeits, learning models which are faster and make fewer mistakes than manual checking, protecting brands, ensuring quality and combating fraud. Counterfeit logos have become a major challenge for industries, e-commerce platforms, and consumers threatening brand integrity, customer trust, and economic growth. This paper design and implement a machine learning based system capable of accurately detecting fake logos, providing an automated solution for brand authentication and digital security. The dataset comprised genuine and counterfeit logo images sourced from open access such as FlickrLogos-32 and augmented using data preprocessing techniques such as rotation, scaling, and noise insertion. Convolutional Neural Networks (CNNs) were employed as the primary classification model, with feature extraction and image segmentation techniques applied to enhance detection accuracy. The system was simulated using Python libraries (TensorFlow and Keras), and performance evaluation was carried out based on precision, recall, and F1-score metrics. The simulation results demonstrate the robustness of CNNs in handling variations in color, shape, and resolution, thereby validating their suitability for real-world applications. Findings reveal that the model achieved a classification accuracy of 94.7%, with strong precision and recall rates in distinguishing between authentic and counterfeit logos. The system offers a scalable solution for e-commerce platforms and provides practical implications for policymakers and industries in reducing counterfeit trade.

• FJET_12_95_88

Copyright (c) 2025 FUDMA Journal of Engineering and Technology

This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.